Applied AI, AutoML & MLOps
—
from Research to Production.

Research &
Technical Interests

AI Systems Design
Representation Learning
Graphs & Knowledge Representation
Learning System Evaluation
Production ML Systems

Selected technologies
Python, PyTorch, scikit-learn, Transformers, Docker

I design and evaluate AI systems for complex, heterogeneous and multimodal problems. My work spans representation learning, graph-based knowledge structures, LLM systems, model benchmarking, and the engineering practices required to make ML experimentation reproducible and deployment scalable.

I am particularly interested in how complex problems can be represented, learned and solved through the interaction of multiple intelligent components.

ABOUT THIS WEBSITE

This website is a portfolio of my work, experiments and technical thinking around artificial intelligence and machine learning.

It brings together research notes, engineering projects, case studies, technical articles and educational resources covering topics from representation learning and graph-based AI to ML systems and production engineering.

Building AI systems

AI System Design

Designing AI systems as interconnected components, from problem framing and architectural choices to the integration of models, data, knowledge and decision mechanisms.

Representation & Knowledge

Building meaningful representations of data and knowledge using embeddings, multimodal representations, graphs and knowledge structures to capture the information and relationships required by the problem.

Model Design & Selection

Designing, adapting and selecting learning models through systematic benchmarking, optimization and AutoML methods, according to performance, computational and operational constraints.

Evaluation & Experimentation

Designing reproducible experimental frameworks, metrics, benchmarks and ablation studies to evaluate hypotheses, compare approaches and produce reliable evidence for technical decisions.

MLOps & Production

Turning experimental AI systems into reproducible, scalable and maintainable production systems through automated pipelines, lifecycle management, deployment and observability.

Recent Post

You will find here the latest posts published in the Blog section.

14/03/23

Fonctions d'activation: Partie 3

Cette dernière partie aborde plusieurs alternatives de ReLU, leur propriétés mathématiques, leurs avantages, leurs limitations et leurs cas d'utilisation.

  • Deep Learning
  • Mathematics
11/03/23

Fonctions d'activation: Partie 2

Cette partie explore les différentes fonctions de type rectifier dérivées de ReLU. Ces fonctions figurent parmi les activations les plus couramment utilisées en Deep Learning.

  • Deep Learning
  • Mathematics
09/03/23

Fonctions d'activation: Partie 1

Dans cette partie: Explorez les avantages et limites des fonction d'activations fondamentales, ainsi que leur lien avec les neurones biologiques.

  • Deep Learning
  • Mathematics
  • Fundamentals

Interested in my work?

Resume & contact
LLM Research Engineer
CV AI Systems Architect
CV LinkedIn